performance-direction
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill references multiple external URLs for technical documentation and industry guidelines. These links target trusted organizations (e.g., Microsoft, Meta, Netflix, OpenAI, SAG-AFTRA) and well-known service providers (e.g., ElevenLabs, Synthesia, HeyGen). All references are for educational and configuration purposes and lead to official domains.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided scripts, character descriptions, and scene requirements to generate performance briefs. This ingestion of untrusted data represents a potential surface for indirect prompt injection; however, the skill mitigates this by providing rigid structured handoff formats, specific beat-mapping instructions, and explicit boundaries for the AI's output.
- [SAFE]: The skill emphasizes best practices for digital rights management, requiring confirmation of consent and authorized likeness usage before directing synthetic personas. It explicitly instructs the agent to refuse requests involving non-consensual soundalikes or lookalikes.
Audit Metadata